Click here to close now.


Apache Authors: Jim Scott, Liz McMillan, Craig Lowell, AppDynamics Blog, Dana Gardner

Related Topics: Java IoT, Microservices Expo, Adobe Flex, IoT User Interface, Apache

Java IoT: Article

Why Averages Are Inadequate, and Percentiles Are Great

Averages are ineffective because they are too simplistic and one-dimensional

Anyone who ever monitored or analyzed an application uses or has used averages. They are simple to understand and calculate. We tend to ignore just how wrong the picture is that averages paint of the world. To emphasis the point let me give you a real-world example outside of the performance space that I read recently in a newspaper.

The article was explaining that the average salary in a certain region in Europe was 1900 Euro's (to be clear this would be quite good in that region!). However when looking closer they found out that the majority, namely 9 out of 10 people, only earned around 1000 Euros and one would earn 10.000 (I over simplified this of course, but you get the idea). If you do the math you will see that the average of this is indeed 1900, but we can all agree that this does not represent the "average" salary as we would use the word in day to day live. So now let's apply this thinking to application performance.

The Average Response Time
The average response time is by far the most commonly used metric in application performance management. We assume that this represents a "normal" transaction, however this would only be true if the response time is always the same (all transaction run at equal speed) or the response time distribution is roughly bell curved.

A Bell curve represents the "normal" distribution of response times in which the average and the median are the same. It rarely ever occurs in real applications

In a Bell Curve the average (mean) and median are the same. In other words observed performance would represent the majority (half or more than half) of the transactions.

In reality most applications have few very heavy outliers; a statistician would say that the curve has a long tail. A long tail does not imply many slow transactions, but few that are magnitudes slower than the norm.

This is a typical Response Time Distribution with few but heavy outliers - it has a long tail. The average here is dragged to the right by the long tail.

We recognize that the average no longer represents the bulk of the transactions but can be a lot higher than the median.

You can now argue that this is not a problem as long as the average doesn't look better than the median. I would disagree, but let's look at another real-world scenario experienced by many of our customers:

This is another typical Response Time Distribution. Here we have quite a few very fast transactions that drag the average to the left of the actual median

In this case a considerable percentage of transactions are very, very fast (10-20 percent), while the bulk of transactions are several times slower. The median would still tell us the true story, but the average all of a sudden looks a lot faster than most of our transactions actually are. This is very typical in search engines or when caches are involved - some transactions are very fast, but the bulk are normal. Another reason for this scenario are failed transactions, more specifically transactions that failed fast. Many real-world applications have a failure rate of 1-10 percent (due to user errors or validation errors). These failed transactions are often magnitudes faster than the real ones and consequently distorted an average.

Of course performance analysts are not stupid and regularly try to compensate with higher frequency charts (compensating by looking at smaller aggregates visually) and by taking in minimum and maximum observed response times. However we can often only do this if we know the application very well, those unfamiliar with the application might easily misinterpret the charts. Because of the depth and type of knowledge required for this, it's difficult to communicate your analysis to other people - think how many arguments between IT teams have been caused by this. And that's before we even begin to think about communicating with business stakeholders!

A better metric by far are percentiles, because they allow us to understand the distribution. But before we look at percentiles, let's take a look a key feature in every production monitoring solution: Automatic Baselining and Alerting.

Automatic Baselining and Alerting
In real-world environments, performance gets attention when it is poor and has a negative impact on the business and users. But how can we identify performance issues quickly to prevent negative effects? We cannot alert on every slow transaction, since there are always some. In addition, most operations teams have to maintain a large number of applications and are not familiar with all of them, so manually setting thresholds can be inaccurate, quite painful and time-consuming.

The industry has come up with a solution called Automatic Baselining. Baselining calculates out the "normal" performance and only alerts us when an application slows down or produces more errors than usual. Most approaches rely on averages and standard deviations.

Without going into statistical details, this approach again assumes that the response times are distributed over a bell curve:

The Standard Deviation represents 33% of all transactions with the mean as the middle. 2xStandard Deviation represents 66% and thus the majority, everything outside could be considered an outlier. However most real world scenarios are not bell curved...

Typically, transactions that are outside two times standard deviation are treated as slow and captured for analysis. An alert is raised if the average moves significantly. In a bell curve this would account for the slowest 16.5 percent (and you can of course adjust that); however; if the response time distribution does not represent a bell curve, it becomes inaccurate. We either end up with a lot of false positives (transactions that are a lot slower than the average but when looking at the curve lie within the norm) or we miss a lot of problems (false negatives). In addition if the curve is not a bell curve, then the average can differ a lot from the median; applying a standard deviation to such an average can lead to quite a different result than you would expect. To work around this problem these algorithms have many tunable variables and a lot of "hacks" for specific use cases.

Why I Love Percentiles
A percentile tells me which part of the curve I am looking at and how many transactions are represented by that metric. To visualize this look at the following chart:

This chart shows the 50th and 90th percentile along with the average of the same transaction. It shows that the average is influenced far mor heavily by the 90th, thus by outliers and not by the bulk of the transactions

The green line represents the average. As you can see it is very volatile. The other two lines represent the 50th and 90th percentile. As we can see the 50th percentile (or median) is rather stable but has a couple of jumps. These jumps represent real performance degradation for the majority (50%) of the transactions. The 90th percentile (this is the start of the "tail") is a lot more volatile, which means that the outliers slowness depends on data or user behavior. What's important here is that the average is heavily influenced (dragged) by the 90th percentile, the tail, rather than the bulk of the transactions.

If the 50th percentile (median) of a response time is 500ms that means that 50% of my transactions are either as fast or faster than 500ms. If the 90th percentile of the same transaction is at 1000ms it means that 90% are as fast or faster and only 10% are slower. The average in this case could either be lower than 500ms (on a heavy front curve), a lot higher (long tail) or somewhere in between. A percentile gives me a much better sense of my real world performance, because it shows me a slice of my response time curve.

For exactly that reason percentiles are perfect for automatic baselining. If the 50th percentile moves from 500ms to 600ms I know that 50% of my transactions suffered a 20% performance degradation. You need to react to that.

In many cases we see that the 75th or 90th percentile does not change at all in such a scenario. This means the slow transactions didn't get any slower, only the normal ones did. Depending on how long your tail is the average might not have moved at all in such a scenario.!

In other cases we see the 98th percentile degrading from 1s to 1.5 seconds while the 95th is stable at 900ms. This means that your application as a whole is stable, but a few outliers got worse, nothing to worry about immediately. Percentile-based alerts do not suffer from false positives, are a lot less volatile and don't miss any important performance degradations! Consequently a baselining approach that uses percentiles does not require a lot of tuning variables to work effectively.

The screenshot below shows the Median (50th Percentile) for a particular transaction jumping from about 50ms to about 500ms and triggering an alert as it is significantly above the calculated baseline (green line). The chart labeled "Slow Response Time" on the other hand shows the 90th percentile for the same transaction. These "outliers" also show an increase in response time but not significant enough to trigger an alert.

Here we see an automatic baselining dashboard with a violation at the 50th percentile. The violation is quite clear, at the same time the 90th percentile (right upper chart) does not violate. Because the outliers are so much slower than the bulk of the transaction an average would have been influenced by them and would not have have reacted quite as dramatically as the 50th percentile. We might have missed this clear violation!

How Can We Use Percentiles for Tuning?
Percentiles are also great for tuning, and giving your optimizations a particular goal. Let's say that something within my application is too slow in general and I need to make it faster. In this case I want to focus on bringing down the 90th percentile. This would ensure sure that the overall response time of the application goes down. In other cases I have unacceptably long outliers I want to focus on bringing down response time for transactions beyond the 98th or 99th percentile (only outliers). We see a lot of applications that have perfectly acceptable performance for the 90th percentile, with the 98th percentile being magnitudes worse.

In throughput oriented applications on the other hand I would want to make the majority of my transactions very fast, while accepting that an optimization makes a few outliers slower. I might therefore make sure that the 75th percentile goes down while trying to keep the 90th percentile stable or not getting a lot worse.

I could not make the same kind of observations with averages, minimum and maximum, but with percentiles they are very easy indeed.

Averages are ineffective because they are too simplistic and one-dimensional. Percentiles are a really great and easy way of understanding the real performance characteristics of your application. They also provide a great basis for automatic baselining, behavioral learning and optimizing your application with a proper focus. In short, percentiles are great!

More Stories By Michael Kopp

Michael Kopp has over 12 years of experience as an architect and developer in the Enterprise Java space. Before coming to CompuwareAPM dynaTrace he was the Chief Architect at GoldenSource, a major player in the EDM space. In 2009 he joined dynaTrace as a technology strategist in the center of excellence. He specializes application performance management in large scale production environments with special focus on virtualized and cloud environments. His current focus is how to effectively leverage BigData Solutions and how these technologies impact and change the application landscape.

Comments (1) View Comments

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.

Most Recent Comments
rtalexander 11/21/12 12:58:00 AM EST

Hey, could you post a reference or two that covers the theory and/or practicalities of the approach you describe?


@ThingsExpo Stories
"Matrix is an ambitious open standard and implementation that's set up to break down the fragmentation problems that exist in IP messaging and VoIP communication," explained John Woolf, Technical Evangelist at Matrix, in this interview at @ThingsExpo, held Nov 4–6, 2014, at the Santa Clara Convention Center in Santa Clara, CA.
WebRTC has had a real tough three or four years, and so have those working with it. Only a few short years ago, the development world were excited about WebRTC and proclaiming how awesome it was. You might have played with the technology a couple of years ago, only to find the extra infrastructure requirements were painful to implement and poorly documented. This probably left a bitter taste in your mouth, especially when things went wrong.
Nowadays, a large number of sensors and devices are connected to the network. Leading-edge IoT technologies integrate various types of sensor data to create a new value for several business decision scenarios. The transparent cloud is a model of a new IoT emergence service platform. Many service providers store and access various types of sensor data in order to create and find out new business values by integrating such data.
The broad selection of hardware, the rapid evolution of operating systems and the time-to-market for mobile apps has been so rapid that new challenges for developers and engineers arise every day. Security, testing, hosting, and other metrics have to be considered through the process. In his session at Big Data Expo, Walter Maguire, Chief Field Technologist, HP Big Data Group, at Hewlett-Packard, will discuss the challenges faced by developers and a composite Big Data applications builder, focusing on how to help solve the problems that developers are continuously battling.
There are so many tools and techniques for data analytics that even for a data scientist the choices, possible systems, and even the types of data can be daunting. In his session at @ThingsExpo, Chris Harrold, Global CTO for Big Data Solutions for EMC Corporation, will show how to perform a simple, but meaningful analysis of social sentiment data using freely available tools that take only minutes to download and install. Participants will get the download information, scripts, and complete end-to-end walkthrough of the analysis from start to finish. Participants will also be given the pract...
WebRTC: together these advances have created a perfect storm of technologies that are disrupting and transforming classic communications models and ecosystems. In his session at WebRTC Summit, Cary Bran, VP of Innovation and New Ventures at Plantronics and PLT Labs, will provide an overview of this technological shift, including associated business and consumer communications impacts, and opportunities it may enable, complement or entirely transform.
SYS-CON Events announced today that Dyn, the worldwide leader in Internet Performance, will exhibit at SYS-CON's 17th International Cloud Expo®, which will take place on November 3-5, 2015, at the Santa Clara Convention Center in Santa Clara, CA. Dyn is a cloud-based Internet Performance company. Dyn helps companies monitor, control, and optimize online infrastructure for an exceptional end-user experience. Through a world-class network and unrivaled, objective intelligence into Internet conditions, Dyn ensures traffic gets delivered faster, safer, and more reliably than ever.
WebRTC services have already permeated corporate communications in the form of videoconferencing solutions. However, WebRTC has the potential of going beyond and catalyzing a new class of services providing more than calls with capabilities such as mass-scale real-time media broadcasting, enriched and augmented video, person-to-machine and machine-to-machine communications. In his session at @ThingsExpo, Luis Lopez, CEO of Kurento, will introduce the technologies required for implementing these ideas and some early experiments performed in the Kurento open source software community in areas ...
Too often with compelling new technologies market participants become overly enamored with that attractiveness of the technology and neglect underlying business drivers. This tendency, what some call the “newest shiny object syndrome,” is understandable given that virtually all of us are heavily engaged in technology. But it is also mistaken. Without concrete business cases driving its deployment, IoT, like many other technologies before it, will fade into obscurity.
Today air travel is a minefield of delays, hassles and customer disappointment. Airlines struggle to revitalize the experience. GE and M2Mi will demonstrate practical examples of how IoT solutions are helping airlines bring back personalization, reduce trip time and improve reliability. In their session at @ThingsExpo, Shyam Varan Nath, Principal Architect with GE, and Dr. Sarah Cooper, M2Mi's VP Business Development and Engineering, will explore the IoT cloud-based platform technologies driving this change including privacy controls, data transparency and integration of real time context w...
Who are you? How do you introduce yourself? Do you use a name, or do you greet a friend by the last four digits of his social security number? Assuming you don’t, why are we content to associate our identity with 10 random digits assigned by our phone company? Identity is an issue that affects everyone, but as individuals we don’t spend a lot of time thinking about it. In his session at @ThingsExpo, Ben Klang, Founder & President of Mojo Lingo, will discuss the impact of technology on identity. Should we federate, or not? How should identity be secured? Who owns the identity? How is identity ...
The IoT market is on track to hit $7.1 trillion in 2020. The reality is that only a handful of companies are ready for this massive demand. There are a lot of barriers, paint points, traps, and hidden roadblocks. How can we deal with these issues and challenges? The paradigm has changed. Old-style ad-hoc trial-and-error ways will certainly lead you to the dead end. What is mandatory is an overarching and adaptive approach to effectively handle the rapid changes and exponential growth.
The buzz continues for cloud, data analytics and the Internet of Things (IoT) and their collective impact across all industries. But a new conversation is emerging - how do companies use industry disruption and technology enablers to lead in markets undergoing change, uncertainty and ambiguity? Organizations of all sizes need to evolve and transform, often under massive pressure, as industry lines blur and merge and traditional business models are assaulted and turned upside down. In this new data-driven world, marketplaces reign supreme while interoperability, APIs and applications deliver un...
Electric power utilities face relentless pressure on their financial performance, and reducing distribution grid losses is one of the last untapped opportunities to meet their business goals. Combining IoT-enabled sensors and cloud-based data analytics, utilities now are able to find, quantify and reduce losses faster – and with a smaller IT footprint. Solutions exist using Internet-enabled sensors deployed temporarily at strategic locations within the distribution grid to measure actual line loads.
The Internet of Everything is re-shaping technology trends–moving away from “request/response” architecture to an “always-on” Streaming Web where data is in constant motion and secure, reliable communication is an absolute necessity. As more and more THINGS go online, the challenges that developers will need to address will only increase exponentially. In his session at @ThingsExpo, Todd Greene, Founder & CEO of PubNub, will explore the current state of IoT connectivity and review key trends and technology requirements that will drive the Internet of Things from hype to reality.
The Internet of Things (IoT) is growing rapidly by extending current technologies, products and networks. By 2020, Cisco estimates there will be 50 billion connected devices. Gartner has forecast revenues of over $300 billion, just to IoT suppliers. Now is the time to figure out how you’ll make money – not just create innovative products. With hundreds of new products and companies jumping into the IoT fray every month, there’s no shortage of innovation. Despite this, McKinsey/VisionMobile data shows "less than 10 percent of IoT developers are making enough to support a reasonably sized team....
You have your devices and your data, but what about the rest of your Internet of Things story? Two popular classes of technologies that nicely handle the Big Data analytics for Internet of Things are Apache Hadoop and NoSQL. Hadoop is designed for parallelizing analytical work across many servers and is ideal for the massive data volumes you create with IoT devices. NoSQL databases such as Apache HBase are ideal for storing and retrieving IoT data as “time series data.”
Today’s connected world is moving from devices towards things, what this means is that by using increasingly low cost sensors embedded in devices we can create many new use cases. These span across use cases in cities, vehicles, home, offices, factories, retail environments, worksites, health, logistics, and health. These use cases rely on ubiquitous connectivity and generate massive amounts of data at scale. These technologies enable new business opportunities, ways to optimize and automate, along with new ways to engage with users.
The IoT is upon us, but today’s databases, built on 30-year-old math, require multiple platforms to create a single solution. Data demands of the IoT require Big Data systems that can handle ingest, transactions and analytics concurrently adapting to varied situations as they occur, with speed at scale. In his session at @ThingsExpo, Chad Jones, chief strategy officer at Deep Information Sciences, will look differently at IoT data so enterprises can fully leverage their IoT potential. He’ll share tips on how to speed up business initiatives, harness Big Data and remain one step ahead by apply...
There will be 20 billion IoT devices connected to the Internet soon. What if we could control these devices with our voice, mind, or gestures? What if we could teach these devices how to talk to each other? What if these devices could learn how to interact with us (and each other) to make our lives better? What if Jarvis was real? How can I gain these super powers? In his session at 17th Cloud Expo, Chris Matthieu, co-founder and CTO of Octoblu, will show you!